Knowledge Lens is an India-based startup that builds domain-specific Data Lenses that enable the discovery of actionable insights from Enterprise Data Assets. Our team of tech experts work on cutting edge technologies such as Big Data, AI, IoT, Blockchain and Cloud to add differentiated value to our customers. Our mission is to turn enterprises’ dark data into meaningful business insights. Here is an example of how we helped solve a challenge for one of our customers.
Client Introduction & Business Requirement:
Our client is a Global Logistics Company based out of India.
Catering to increased demand in online retail, modern warehouses today faces many challenges from sorting, tracking and transit operations to seamless delivery. Security risks are one of the biggest and most commonly faced problems as these warehouses are usually setup in the outskirts of the city with proximity to transit hubs such as railways and airports. Warehouses have multiple entry and exit points, which are vulnerable to frequent attacks from trespassers, leading to financial losses for the company. Industry statistics show that warehouse locations remained as the most commonplace for cargo thefts. In order to avoid security threats like burglary, theft and vandalism, logistics companies today are looking for highly advanced, technology-driven state-of-the-art surveillance infrastructure for their warehouses.
Our client was seeking a means to improve the security for many of their warehouses . We were able to deliver a highly scalable and robust solution to them with our Perimeter Intrusion Detection System (PIDS) – an end-to-end solution that includes:
- Detecting persons attempting to gain unauthorized entry into the premises and identifying length of unauthorized stay
- Capturing the image of the intruder
- Identifying suspicious activities (odd hours, scaling walls, unauthorized package removal etc) around the warehouse premise.
Solution Overview:
Perimeter Intrusion detection is essential in protecting critical assets and resources in a warehouse. In the case of detecting unauthorized physical access, the algorithm is fed the video stream directly from the source and it detects and raises an alert about the presence of any suspicious entity. The solution detects intrusion and raises an alarm on suspicious activities, to notify the security guard or command center.
A unique combination of Intel’s powerful hardware and the latest AI driven innovation helped us build a robust solution for our customer. We started by planning to optimize cost by efficiently using the hardware and determining how to address the scalability issues. While working with the Intel AI Builders technical team, we developed a compatible AI model architecture and used the Intel OpenVINO toolkit to optimize those resource hungry models. With the help of the Intel team, we were able to leverage the quantization capabilities available via DL Boost on 2nd Gen Intel Xeon Scalable processors. These optimizations, along with the accuracy-aware training that we did, allowed us to significantly improve the inference performance of our models, without a loss in accuracy.
We are now able to run largee number of models in parallel on the same underlying hardware infrastructure, which means our architecture can now support up to 30 cameras. The approach has helped us optimize costs for the customers, while ensuring efficient hardware utilization and being able to scale the number of AI models.
The benefits we have been able to deliver include:
- Reduced processor space consumption by 20-30%
- Runs up to 30 parallel streams on a single machine.
- Cost reduction by 50-60% on hardware used.
- Core usage of AI models retained, without losing accuracy.
At Knowledge Lens, we constantly work towards improving our product technologies, so that our customers can focus on running their business efficiently. We have partnered with Intel to bring AI to the center of your data-centric strategy and help business leaders augment threat detection capabilities.
Intel is committed to respecting human rights and avoiding complicity in human rights abuses. See Intel's Global Human Rights Principles. Intel's products and software are intended only to be used in applications that do not cause or contribute to a violation of an internationally recognized human right.